refactor(clinical_report): clean id signatures

.github/workflows/ci.yaml

@@ -25,7 +25,7 @@ jobs:
         run: pip install uv
 
       - name: Install dependencies
-        run: uv sync --group dev
+        run: uv sync --locked --group dev
 
       - name: Run pre-commit (lint + format + tests)
-        run: uv run pre-commit run --all-files
+        run: uv run pre-commit run --all-files

pyproject.toml

@@ -20,7 +20,6 @@ dependencies = [
     "connectorx>=0.4.3",
     "fastexcel>=0.16.0",
     "httpx>=0.28.0",
-    "python-dotenv>=1.0.0",
     "hydra-core>=1.3.2",
     "loguru>=0.7.3",
     "ontoma[ner]>=2.4.1",

src/mira/provider/chembl/curation.py

@@ -81,7 +81,7 @@ def extract_chembl_ct_curation(
             host=db_host,
             port=db_port,
             service=db_service,
-            db_schema="CHEMBL_36",
+            db_schema="CHEMBL_37",
             limit=None,
             init_client_lib_dir=oracle_client_path,
         )

src/mira/provider/chembl/indications.py

@@ -1,6 +1,7 @@
 """Extraction of clinical reports from ChEMBL drug/indications dataset."""
 
 import polars as pl
+import polars_hash as plh
 
 from mira.dataset import ClinicalReport
 from mira.dataset.clinical_report import APPROVAL_SOURCES
@@ -42,7 +43,7 @@ def extract_clinical_report(
             id=(
                 pl.when(pl.col("ref_type") == "INN")
                 .then(
-                    pl.concat_str(
+                    plh.concat_str(
                         pl.col("ref_id"),
                         pl.lit("/"),
                         pl.col("efo_term"),
@@ -92,6 +93,15 @@ def extract_clinical_report(
             type=pl.lit(ClinicalReportType.INDICATION.value),
         )
         .explode("id")
+        .with_columns(
+            id=(
+                # ID is hashed when it does not relate to the representation in the primary source
+                pl.when(pl.col("source").is_in(["INN", "FDA", "USAN"]))
+                .then(pl.col("id").chash.sha2_256())
+                # For DalyMed, EMA, ATC - ID remains the original value (can be queried in the primary sources)
+                .otherwise(pl.col("id"))
+            )
+        )
         .unique()
     )
 

src/mira/provider/ttd.py

@@ -1,4 +1,5 @@
 import polars as pl
+import polars_hash as plh
 
 from mira.dataset.clinical_report import ClinicalReport
 from mira.schemas import (
@@ -69,9 +70,10 @@ def extract_clinical_report(
 ) -> ClinicalReport:
     """Extract clinical reports from TTD drug/disease dataset."""
     reports = indications.select(
-        id=pl.concat_str(
-            [pl.col("ttd_id"), pl.lit("/"), pl.col("diseaseFromSource")]
-        ).str.to_lowercase(),
+        id=plh.concat_str(
+            pl.col("ttd_id"),
+            pl.col("diseaseFromSource").str.to_lowercase(),
+        ).chash.sha2_256(),
         origin=pl.lit(ClinicalReportOrigin.CURATED_RESOURCE),
         url=pl.concat_str(
             [pl.lit("https://ttd.idrblab.cn/data/drug/details/"), pl.col("ttd_id")]

src/mira/utils/db.py

@@ -1,3 +1,5 @@
+from urllib.parse import quote
+
 import polars as pl
 from loguru import logger
 
@@ -16,10 +18,14 @@ def construct_db_uri(
     db_uri: str,
     db_user: str | None = None,
     db_password: str | None = None,
-):
-    """Constructs a database URI from the given parameters."""
+) -> str:
+    """Construct a database URI, requiring complete credentials when supplied."""
+    if bool(db_user) != bool(db_password):
+        raise ValueError("db_user and db_password must be set together")
     if db_user and db_password:
-        return f"{db_type}://{db_user}:{db_password}@{db_uri}"
+        user = quote(db_user, safe="")
+        password = quote(db_password, safe="")
+        return f"{db_type}://{user}:{password}@{db_uri}"
     return f"{db_type}://{db_uri}"
 
 
@@ -39,7 +45,8 @@ def _build_select_query(
         select_cols: List of columns or a raw
         limit: Optional row limit.
         dialect: "generic" uses SQL LIMIT. "oracle" uses FETCH FIRST.
-        where_clause: Optional WHERE clause (without the WHERE keyword).
+        where_clause: Optional raw SQL WHERE clause (without the WHERE keyword).
+            It is intentionally interpolated as supplied by the caller.
 
     Returns:
         SQL query string.
@@ -56,6 +63,7 @@ def _build_select_query(
         query += f" WHERE {where_clause}"
 
     if limit is not None:
+        limit = int(limit)
         if dialect.lower() == "oracle":
             query += f" FETCH FIRST {limit} ROWS ONLY"
         else:

tests/test_db.py

@@ -0,0 +1,38 @@
+import pytest
+
+from mira.utils.db import _build_select_query, construct_db_uri
+
+
+def test_construct_db_uri_quotes_credentials() -> None:
+    assert (
+        construct_db_uri(
+            "postgresql",
+            "localhost:5432/db",
+            db_user="user@example",
+            db_password="p@ss:/word",
+        )
+        == "postgresql://user%40example:p%40ss%3A%2Fword@localhost:5432/db"
+    )
+
+
+@pytest.mark.parametrize(
+    ("db_user", "db_password"),
+    [("user", None), (None, "password"), ("", "password"), ("user", "")],
+)
+def test_construct_db_uri_rejects_partial_credentials(
+    db_user: str | None, db_password: str | None
+) -> None:
+    with pytest.raises(ValueError, match="set together"):
+        construct_db_uri("postgresql", "localhost/db", db_user, db_password)
+
+
+def test_build_select_query_coerces_limit() -> None:
+    assert _build_select_query("table", "schema", limit="5") == (
+        "SELECT DISTINCT * FROM schema.table LIMIT 5"
+    )
+
+
+def test_build_select_query_documents_raw_where_clause() -> None:
+    assert _build_select_query("table", "schema", where_clause="id > 1") == (
+        "SELECT DISTINCT * FROM schema.table WHERE id > 1"
+    )